nessai.livepoint#

Functions related to creating live points and converting to other common data-types.

Functions#

`add_extra_parameters_to_live_points`(parameters[, ...])	Add extra parameters to the live points dtype.
`reset_extra_live_points_parameters`()	Reset the extra live points parameters.
`get_dtype`(names[, array_dtype, non_sampling_parameters])	Get a list of tuples containing the dtypes for the structured array
`empty_structured_array`(n[, names, dtype, ...])	Get an empty structured array with the extra parameters initialised.
`live_points_to_array`(live_points[, names, copy])	Converts live points to unstructured arrays for training.
`parameters_to_live_point`(parameters, names[, ...])	Take a list or array of parameters for a single live point
`numpy_array_to_live_points`(array, names[, ...])	Convert a numpy array to a numpy structure array with the correct fields
`dict_to_live_points`(d[, non_sampling_parameters])	Convert a dictionary with parameters names as keys to live points.
`live_points_to_dict`(live_points[, names])	Convert a structured array of live points to a dictionary with
`dataframe_to_live_points`(df[, non_sampling_parameters])	Convert and pandas dataframe to live points.
`unstructured_view`(x[, names, dtype])	Get an unstructured view of a live points containing certain parameters.

Module Contents#

nessai.livepoint.add_extra_parameters_to_live_points(parameters, default_values=None)#

Add extra parameters to the live points dtype.

Extra parameters will be included in the live points dtype that is used for constructing/converting to/from live points.

Parameters:

parameters: list: List of parameters to add.
default_values: Optional[Union[List, Tuple]]: List of default values for each parameters. If not specified, default values will be set to based on :code: DEFAULT_FLOAT_VALUE in nessai.config.

nessai.livepoint.reset_extra_live_points_parameters()#: Reset the extra live points parameters.

nessai.livepoint.get_dtype(names, array_dtype=None, non_sampling_parameters=True)#

Get a list of tuples containing the dtypes for the structured array

Parameters:

nameslist of str: Names of parameters
array_dtypeOptional[str]: dtype to use
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

numpy.dtype: A instance of numpy.dtype.

nessai.livepoint.empty_structured_array(n, names=None, dtype=None, non_sampling_parameters=True)#

Get an empty structured array with the extra parameters initialised.

Parameters:

nint: Length of the structured array
dtypeOptional[list]: Dtype to use. Must contain the non-sampling parameters.
namesOptional[list]: Names of fields (excluding non-sampling parameters) to construct the dtype. Must be specified if dtype is not specified.
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

np.ndarray: Structured array with the all parameters initialised to their default values.

nessai.livepoint.live_points_to_array(live_points, names=None, copy=False)#

Converts live points to unstructured arrays for training.

Parameters:

live_pointsstructured_array: Structured array of live points
nameslist of str or None: If None all fields in the structured array are added to the dictionary else only those included in the list are added.
copybool: If true, returns a copy. If false, returns a view. See numpy documentation for numpy.lib.recfunctions.structured_to_unstructured for more details.

Returns:

np.ndarray: Unstructured numpy array

nessai.livepoint.parameters_to_live_point(parameters, names, non_sampling_parameters=True)#

Take a list or array of parameters for a single live point and converts them to a live point.

Returns an empty array with the correct fields if len(parameters) is zero

Parameters:

parameterstuple: Float point values for each parameter
namestuple: Names for each parameter as strings
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

structured_array: Numpy structured array with fields given by names plus logP and logL

nessai.livepoint.numpy_array_to_live_points(array, names, non_sampling_parameters=True)#

Convert a numpy array to a numpy structure array with the correct fields

Parameters:

arraynp.ndarray: Instance of np.ndarray to convert to a structured array
namestuple: Names for each parameter as strings
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

structured_array: Numpy structured array with fields given by names plus logP and logL

nessai.livepoint.dict_to_live_points(d, non_sampling_parameters=True)#

Convert a dictionary with parameters names as keys to live points.

Assumes all entries have the same length. Also, determines if points are array-like by checking if the first value has __len__ attribute and the number of dimensions is not zero, if not the dictionary is assumed to contain a single live point.

Parameters:

ddict: Dictionary with parameters names as keys and values that correspond to one or more parameters
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

structured_array: Numpy structured array with fields given by names plus logP and logL

nessai.livepoint.live_points_to_dict(live_points, names=None)#

Convert a structured array of live points to a dictionary with a key per field.

Parameters:

live_pointsstructured_array: Array of live points
nameslist of str or None: If None all fields in the structured array are added to the dictionary else only those included in the list are added.

Returns:

dict: Dictionary of live points

nessai.livepoint.dataframe_to_live_points(df, non_sampling_parameters=True)#

Convert and pandas dataframe to live points.

Adds the non-sampling parameter initialised to their defaults.

Based on this answer on Stack Exchange: https://stackoverflow.com/a/51280608

Parameters:

dfpandas.DataFrame: Pandas DataFrame to convert to live points
non_sampling_parametersbool: Indicates whether non-sampling parameters should be included.

Returns:

structured_array: Numpy structured array with fields given by column names plus logP and logL.

nessai.livepoint.unstructured_view(x, names=None, dtype=None)#

Get an unstructured view of a live points containing certain parameters.

This is quicker than converting to a unstructured array and does not create a copy of the array.

Parameters:

xnumpy.ndarray: Structured array.
namesOptional[Iterable]: Iterable of parameters to include in the view. Must be specified if dtype is None.
dtypeOptional[numpy.dtype]: Dtype for constructing the unstructured view.

Returns:

numpy.ndarray: View of x as an unstructured array that contains only the parameters in names/dtype. Shape is (x.size, # parameters).